TWON social media: a scalable MERN-Stack platform for experimental research in online social networks

Online social networks play a central role in shaping public discourse, yet con- ducting controlled experimental research on such platforms remains challenging due to limited access, lack of transparency in ranking algorithms, and restricted intervention capabilities. This paper presents TWON, a scalable and modular social media platform designed to enable controlled, reproducible experimen- tation on user behavior, information diffusion, and algorithmic interventions. The platform further incorporates large language model (LLM) capabilities for content generation, moderation, and agent-based simulation, enabling hybrid experimental designs that combine human participants with automated agents. The system is validated through multiple empirical deployments across diverse research contexts, including (i) a disinformation study analyzing user engagement with manipulated news content (272 participants), (ii) a scientific communication study evaluating pre-bunking and uncertainty interventions (1200 participants), (iii) a toxicity prevention study leveraging real-time AI-assisted comment rewrit- ing (574 participants), and (iv) large-scale agent-based simulations exploring conversational dynamics under different ranking strategies (54 agents). Across these studies, TWON supports flexible experimental configurations and cap- tures fine-grained behavioral data, enabling systematic analysis of engagement patterns and intervention effects.